Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH 0/5] Non-fatal repin retries and NUMA-local rebind worker
@ 2026-11-19 11:14 Matthew Auld
  2026-11-19 11:14 ` [PATCH 1/5] drm/xe/vm: restore userptr invalidations on VMA op unwind Matthew Auld
                   ` (6 more replies)
  0 siblings, 7 replies; 22+ messages in thread
From: Matthew Auld @ 2026-11-19 11:14 UTC (permalink / raw)
  To: Matthew Brost, Thomas Hellström, Himal Prasad Ghimiray,
	Stuart Summers
  Cc: intel-xe, Matthew Auld

drm_gpusvm_get_pages() fails with -EBUSY after a fixed 1s budget when
CPU mappings keep changing during a userptr repin (for instance under
AutoNUMA balancing or concurrent mapping churn), which Xe currently
treats as fatal.

Retry userptr repins with locks dropped up to a configurable
userptr_repin_timeout_ms sysfs timeout (requeuing the compute rebind
worker on -EAGAIN), restore userptr invalidations lost on VMA op unwind,
and run the compute rebind worker on the device's NUMA node so HMM
faults do not trigger remote AutoNUMA migrations (hopefully).

First patch is mostly orthogonal bug fix, however series here could
potentially make the race more likely, so included it here.

Entire series is not necessarily intended to be merged upstream as-is.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
---
Matthew Auld (2):
      drm/xe/vm: restore userptr invalidations on VMA op unwind
      drm/xe: run compute rebind worker on device NUMA node

Thomas Hellström (3):
      drm/xe: Requeue the rebind worker instead of retrying in-line on -EAGAIN
      drm/xe: Add a userptr repin timeout sysfs attribute
      drm/xe/userptr: Make the repin retry loop non-fatal

 drivers/gpu/drm/xe/xe_device.c       |  13 +++-
 drivers/gpu/drm/xe/xe_device.h       |   2 +
 drivers/gpu/drm/xe/xe_device_sysfs.c |  56 ++++++++++++++
 drivers/gpu/drm/xe/xe_device_types.h |  18 ++++-
 drivers/gpu/drm/xe/xe_exec.c         |   9 ++-
 drivers/gpu/drm/xe/xe_pagefault.c    |   4 +
 drivers/gpu/drm/xe/xe_userptr.c      |  47 ++++++++++--
 drivers/gpu/drm/xe/xe_userptr.h      |  53 +++++++++++++
 drivers/gpu/drm/xe/xe_vm.c           | 145 ++++++++++++++++++++++++-----------
 drivers/gpu/drm/xe/xe_vm.h           |   6 +-
 drivers/gpu/drm/xe/xe_vm_madvise.c   |  23 +++++-
 drivers/gpu/drm/xe/xe_vm_types.h     |   8 ++
 12 files changed, 328 insertions(+), 56 deletions(-)
---
base-commit: 84508686117ca3be5756049acdf15e3e47fd2683
change-id: 20261118-userptr-repin-52499936a13c


^ permalink raw reply	[flat|nested] 22+ messages in thread

end of thread, other threads:[~2026-11-20 10:50 UTC | newest]

Thread overview: 22+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-19 11:14 [PATCH 0/5] Non-fatal repin retries and NUMA-local rebind worker Matthew Auld
2026-11-19 11:14 ` [PATCH 1/5] drm/xe/vm: restore userptr invalidations on VMA op unwind Matthew Auld
2026-11-19 12:48   ` Thomas Hellström
2026-11-19 16:35   ` Himal prasad Ghimiray
2026-11-19 11:14 ` [PATCH 2/5] drm/xe: Requeue the rebind worker instead of retrying in-line on -EAGAIN Matthew Auld
2026-11-19 12:51   ` Thomas Hellström
2026-11-19 16:29     ` Matthew Auld
2026-11-19 17:26       ` Thomas Hellström
2026-11-19 17:38         ` Matthew Auld
2026-11-19 16:36   ` Himal prasad Ghimiray
2026-11-19 11:14 ` [PATCH 3/5] drm/xe: Add a userptr repin timeout sysfs attribute Matthew Auld
2026-11-19 16:38   ` Himal prasad Ghimiray
2026-11-20 10:43     ` Matthew Auld
2026-11-20 10:49       ` Himal prasad Ghimiray
2026-11-19 11:14 ` [PATCH 4/5] drm/xe/userptr: Make the repin retry loop non-fatal Matthew Auld
2026-11-19 13:03   ` Thomas Hellström
2026-11-19 17:10   ` Himal prasad Ghimiray
2026-11-19 11:14 ` [PATCH 5/5] drm/xe: run compute rebind worker on device NUMA node Matthew Auld
2026-11-19 13:11   ` Thomas Hellström
2026-11-19 13:38     ` Matthew Auld
2026-11-19 13:16 ` ✓ CI.KUnit: success for Non-fatal repin retries and NUMA-local rebind worker Patchwork
2026-11-19 13:54 ` ✓ Xe.CI.BAT: " Patchwork

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox